On the hunt for cherry blossoms

Any day now the miracle of cherry blossoms will begin. Not here in Fernie, of course. We’re still encased in snow and ice. But in Tokyo, beloved cherry blossoms are predicted to appear around March 26 and peak April 2. Meanwhile, in Vancouver, B.C., cherry trees have already started to bloom, evidently, depending on where you look. These gorgeous pale pink puffs of petals are certainly the cheeriest harbingers of springtime. What I didn’t realize, until now, is that they’re also rocking the world of statisticians. Who knew?

Blossoms on tree

Significance, an online and print magazine focused on all things related to statistics and statistical analysis (yup) with articles written by statisticians, recently published an article about When Will The Cherry Trees Bloom In 2025?. I highly recommend the read and not just because then you can claim bragging rights to having read an article by statisticians about statistics.

As it turns out, “bloom date prediction” has real world significance that can impact a variety of concerns including event planning for cherry blossom promoters, harvest scheduling for cherry pickers and climate change study for scientists. But, predicting the appearance of blossoms also has a significant place in statistics history. When statistics was still in its infancy, the Law Of The Flowering Plants, discovered in the mid-1700s, became an important breakthrough in statistics, eventually growing into the huge field of life cycle events study. To this day, predictions like crop yields still use a methodology that is “largely unchanged”. But there’s more!

Flowering tree in spring

Flowering trees at Butchart Gardens on Vancouver Island (both photos above), shot in April, 2023.

The cherry on the top of a statistician’s cake

The study of blossoming cherries, in particular, is being thrown into the spotlight because documenting the timing of the blooms is “among the oldest continuous measurements made by humans, with some bloom date records dating back more than a thousand years. Mind blown.

Celebrating all this awesomeness was the 2025 International Cherry Blossom Prediction Competition. Bloom times in five renowned cherry blossom viewing locations–Washington, D.C., USA, Kyoto, Japan, Vancouver, B.C., Canada, New York City, USA, and Liestal-Weideli, Switzerland–were requested. If you’re getting all excited, relax. The deadline for submissions was in February. This year, the contest had an important twist. Not only were competitors competing against other cherry blossom fans the world over, they were also competing against artificial intelligence (AI). The results are in. You can see all the aggregated predictions (including those submitted by AI) on George Mason University’s website.

Who will be the ultimate winner? I expect anyone who gets a chance to stand under a tree in full cherry blossom petal flurry will think they are the winningest.

2 thoughts on “On the hunt for cherry blossoms

  1. Interesting that there was less than a week’s difference between the human and AI predictions. In one case, there was only one day’s difference!

    This is fascinating – especially the real-world commercial (including touristic) and agricultural implications of this type of forecasting.

    If only this information had been available to the organizers of the Tulip Festival in Ottawa! Our weather was very fickle: some years it was too cold and the flower buds barely opened, other years it was too hot and the flowers got blasted in the heat. It may be that there are larger forces, like whether it is an El Nino or an La Nina year, which also would affect the predictions, and where AI could assist in making more accurate weather predictions. I’m not a meteorologist, so I don’t know, but this seems like a fascinating area to explore.

    Thanks for the article!

    Like

    • You’re so welcome, Martine! I had fun writing it. I know what you mean about wishing other predictions were available but the Significance article hints at why they aren’t that plentiful. The competition required anyone making a prediction to provide an abstract plus “all data and code, along with a Quarto document that reproduces the analysis”. I have no idea what a Quarto document is much less what kind of code you’d need to use but the whole thing sounds really complex. I’ll keep my fingers crossed that this year’s Tulip festival goes off without a hitch!

      Like

Leave a comment